I think my problem can easier be explained via an example: Assume we have a dataset containing the images of 10 different mammals, let's say lion, elephant, cat, ... and horse. We have a 20-class image classification task where we want to detect whether an image belongs to lion-male or lion-female or cat-male or cat-female or ... horse-male or horse-female. As you see, the classes can be simply clustered into 10 mammal classes, i.e., (some) classes are highly dependent.

This example is a kind of made up. Two strategies, imaginable to solve this problem is either we ignore this dependency and treat it as a 20-class classification task or use a hierarchical structure to first classify them into 10 class of mammals and then into male or female. The later however, ignores the fact that there might be (and in fact is) some shared properties across all males or all females that can be used to discriminate female from male.

I was wondering if there is a methodological way to address this type of classification tasks. I'm looking for a keyword to google it. Thanks


1 Answer 1


The type of problem you are describing is known as hierarchical classification, where the topic taxonomy have a structure with some classes being sub-topics of other classes. This presents different challenges in the training and prediction steps ranging from scalability to evaluation metrics.

I would recommend this reference to learn more about this field: Hierarchical Text Classification and Evaluation by Sun and Lim in ICDM '01.

  • $\begingroup$ Here is a software implementation that can model such dependencies via structured output SVM. There are also excellent publications by the software author references on that web page. $\endgroup$
    – Krrr
    Sep 9, 2017 at 8:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.